Primary-receiver metadata availability

Before choosing a prediction error model, establish which independent receiver information is already preserved. This audit uses the same 53 preselected stretches and their saved inspection reports. Group recordings by the existing task/source identifiers; do not select them using model residuals or convergence.

For each recording, verify its filename and original SHA-256 against the stretch catalog. Record the primary-receiver type codes from version strings, NMEA text sentence types and talkers, checksum counts, total stream bytes and bytes not parsed as NMEA. Preserve missing reports and malformed-record failures.

GST/GBS presence is an availability indicator, not validation of a covariance. GSA and GGA quality information do not become noise standard deviations. Unparsed bytes remain unidentified; do not call them raw pseudoranges, receiver precision or evidence of a particular binary protocol. Receiver type codes do not establish manufacturer identity or independent physical instruments.

The study reads no original IMU file, performs no native prediction or fit, and uses no fused-navigation covariance or model outcomes. It cannot establish an aircraft error law or change any model preference.

Freeze the catalog copy, metadata copies, source and environment through the common study runner. Allow one task per distinct catalog recording, at most 53 tasks and 1 MB per inspection input. Use one worker with one numerical-thread allocation and a 120-second cooperative clock allowance. Native calls, packet visits, bound evaluations and optimizer starts are all zero. No retries or budget expansion.

Recompute the inventory from sealed inputs for publication, preserving every file outcome. Archive duplicate source/input/evidence copies with verified aliases. Future checksummed receiver-stream extraction or covariance development require a separate scope; this audit authorizes neither.